Context-Aware Predictive Process Monitoring: The Impact of News Sentiment

Predictive business process monitoring is concerned with forecasting how a process is likely to proceed, covering questions such as what is the next activity to expect and what is the remaining time until case completion. Process prediction typically builds on machine learning techniques that leverage past process execution data. A fundamental problem of a process prediction methods is the data acquisition. So far, research on predictive monitoring utilize data, which is internal to the process. In this paper, we present a novel approach of integrating the external context of the business processes into prediction methods. More specifically, we develop a technique that leverages the sentiments of online news for the task of remaining time prediction. Using our prototypical implementation, we carried out experiments that demonstrate the usefulness of this approach and allowing us to draw conclusions about circumstances in which it works best.